[OLSR-users] Anyone tried two nodes with internet connection?

Rajesh Narayanan (spam-protected)
Thu Jan 11 23:32:48 CET 2007


On 1/11/07, Bernd Petrovitsch <(spam-protected)> wrote:
>
> On Thu, 2007-01-11 at 12:25 -0800, Rajesh Narayanan wrote:
> > :-)
> >
> > This is precisely the question that I have been myself thinking since
> > yesterday. I thought of going back to routing fundamentals to see if
> > there is a good answer. So far I have not been able to think of
>
> Actually Andreas delivered the two possibilities in his answer. One is
> below and the other is: Take dedicated nodes for this.


************ sorry if im starting to sound silly :-), but I didnt understand
this one.


>  anything except a 'not so clean hack' that would sound soemthing like
> > this:
> >
> > 1. Write a script to ping (dyn-gw already does this) on the default
> > WAN interface.
> > 2. If Pings succeed dont do anything
> > 3. If ping fails then delete the default route.
> >     - This is the part you were talking about. HOW DO I RECOVER??
> >     - Since if I delete the route then the pings are not going out of
> > this interface anyways.
>
> The other solution (see Andreas email) is to add (low priority) host
> routes to known-outside ping targets - the ones which are used to decide
> if we have an uplink or not.


**********  this does not help. The issue is with the static route that gets
added when the uplink is discovered. Once its discovered and the static
default route gets added, no amount of dancing with olsrd or the dyn-gw
parameters will be useful.



> 4. The hack is as follows:
> >     - every 5 mins add the default route.
> >     - test ping through that interface
> >     - if successful then leave it
> >     - else, delete the default route again.
> >
> > I know this is BAAAAAAAAAAAD, but all I want is that I should use the
>
> I (also) have some bad feelings about this:
> IMHO this is not very robust (e.g. what happens if olsrd exits somewhere
> in between - it could be a `kill -9` from the outside). Since OLSRD runs
> on embedded hardware, no user action can be expected (let alone
> enforced) so it must work fully automatically (at least in theory we
> have o bugs in implementations;-).
>
> And I assume that OLSRD doesn't announce that link during the test.


********** I just want to prototype this to showcase a possible network
topology so a bad hack is hopefully ok for now. I realize there may be
better ways of doing it. But my coding skills are pretty rusty right now so
scripting is the best choice for me now :-)

Thanks,
Rajesh
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.olsr.org/pipermail/olsr-users/attachments/20070111/b3aed1fe/attachment.html>


More information about the Olsr-users mailing list